Abstract: The quantity of data generated from Twitter and other social networks is enormous and expanding rapidly because of the growing number of users online who share their opinions and thoughts on these platforms. Extracting useful information from these data would be helpful for decision making related to services, products or people. One type of extracting information from these data is Sentiment Analysis (SA) it refers to prediction of the polarity of words to classify the expressed written feelings and opinions into positive or negative. Therefore, SA gives the organizations the ability to observe people’s feelings on particular issue for example their brands and products. Although, a wide range of methods have been deployed to make such analysis but it can be used for Latin texts. On the other hand, the more complex to analyze and morphologically rich Arabic language generate a big sum of data through social media but very few analysis have been conducted on this language and its big variety of dialects. This study surveys the SA of Arabic contents, focusing on the lexicon-based methods used for extracting sentiment from Arabic Tweets written in Modern Standard Arabic (MSA) and dialectical forms. Besides, reviewing Arabic language challenges, along with going through the pre-processing tools used in the literature with some recommendations. Furthermore, showing how they generate sentiment lexicons and how they handled negation.